The Mann-Whitney and Wilcoxon tests are rank sum tests, not median tests. It is quite possible for the ranks to differ but the medians to be the same. This website has a nice example...
http://www.ats.ucla.edu/stat/mult_pkg/faq/general/mann-whitney.htm
The Mann-Whitney and Wilcoxon tests are rank sum tests, not median tests. It is quite possible for the ranks to differ but the medians to be the same. This website has a nice example...
http://www.ats.ucla.edu/stat/mult_pkg/faq/general/mann-whitney.htm
As far as I know, both tests are used for non-parametric data (not normally distributed) or for groups with very low numbers for the first test and larger number of samples for the second. The first test is used to see difference between iwo unmatched samples whereas the second test to see the dufference between two matched samples and larger number of samples. So difference in number of samples sometimes gives same median but sure with different standard error and standard deviation. It is always very wrong to express your data with only mean or median without Standard error or diviation.
Dear Brijesh,
If two groups have an identical or similar distribution of scores shape AND have identical medians then you can be sure that there is no significant difference between them. However, if they have equal medians but different distribution shapes then it is not uncommon to find a significant difference.
- Badges
- Science topic